- Pharo-dev - lists.pharo.org

This week (46/2023) on the Pharo Issue Tracker

by Marcus Denker

# Pharo 11 Backports - backport: Guard RubAbstractTextArea>>#recomputeSelection for empty paragraphs #15140 #15375 https://github.com/pharo-project/pharo/pull/15375 - backport "Defer change list update to avoid a race condition" #14853 #15374 https://github.com/pharo-project/pharo/pull/15374 - 15372-backport-Pharo11-safer-morph-access-in-NECMenuMorphmouseDown-14356 #15373 https://github.com/pharo-project/pharo/pull/15373 # Fixes - Fix scaling of #blankIcon and #notFoundIcon on ThemeIcons #15399 https://github.com/pharo-project/pharo/pull/15399 - 15299-UFFI-should-release-pointers-allocated-for-packing-for-arity #15300 https://github.com/pharo-project/pharo/pull/15300 - 15302-FreeTypeFonts-should-not-use-autorelease-of-ExternalAddresses #15303 https://github.com/pharo-project/pharo/pull/15303 - Calypso: Fix display of extended classes in project view #15314 https://github.com/pharo-project/pharo/pull/15314 - Fix imageCleaner concurrency issue when unifying literals #15385 https://github.com/pharo-project/pharo/pull/15385 - Do not log recompilation in the source file #15398 https://github.com/pharo-project/pharo/pull/15398 - Guard removals of morphs in layouts #1467 https://github.com/pharo-spec/Spec/pull/1467 # Features - Extend ThemeIcons to load the icons for all scales in the icon pack #15323 https://github.com/pharo-project/pharo/pull/15323 - Auto completion should propose globals #15392 https://github.com/pharo-project/pharo/pull/15392 - Calypso: Show content of projects #15315 https://github.com/pharo-project/pharo/pull/15315 # Package Model Cleanup - Packages: Fix missing extensions after class recompiling #15326 https://github.com/pharo-project/pharo/pull/15326 - Introduce a method to get the meta model announcer #15310 https://github.com/pharo-project/pharo/pull/15310 - Packages: Remove #definedSelectors cache #15278 https://github.com/pharo-project/pharo/pull/15278 - Packages: Removing more references to #category #15320 https://github.com/pharo-project/pharo/pull/15320 - RPackage: Deprecate #includesDefinedSelector:ofClass: #15214 https://github.com/pharo-project/pharo/pull/15214 - Package: Move extension methods of RPackage #15346 https://github.com/pharo-project/pharo/pull/15346 - Rename RPackageOrganizer into PackageOrganizer #15360 https://github.com/pharo-project/pharo/pull/15360 - Monticello: Remove usage of RPackageSet #15363 https://github.com/pharo-project/pharo/pull/15363 - Update RPackage dependencies #15393 https://github.com/pharo-project/pharo/pull/15393 - DependencyAnalyzer: Remove usage of RPackageSet #15368 https://github.com/pharo-project/pharo/pull/15368 # Monticello - Remove MCPatchBrowser #15376 https://github.com/pharo-project/pharo/pull/15376 - Monticello: Introduce Monticello-Model #15350 https://github.com/pharo-project/pharo/pull/15350 - Monticello: Simplify method loading #15292 https://github.com/pharo-project/pharo/pull/15292 - Monticello: Remove dead code #15351 https://github.com/pharo-project/pharo/pull/15351 - Monticello: Move mocks to mock package #15361 https://github.com/pharo-project/pharo/pull/15361 - Remove WorkingCopy spotter entries #610 https://github.com/pharo-spec/NewTools/pull/610 # MetaCello - Cleanup metacello #15287 https://github.com/pharo-project/pharo/pull/15287 - Metacello: New cleaning pass #15312 https://github.com/pharo-project/pharo/pull/15312 - Add missing methods for SmalltalkCI #15339 https://github.com/pharo-project/pharo/pull/15339 - Fixes #15342: Monticello: GitHub CI fails with DNU #15344 https://github.com/pharo-project/pharo/pull/15344 - Metacello: fix block evaluation #15362 https://github.com/pharo-project/pharo/pull/15362 # Kernel - Allow to create anonymous classes in a specific environment #15309 https://github.com/pharo-project/pharo/pull/15309 - Fix bug in remake of anonymous classes #15343 https://github.com/pharo-project/pharo/pull/15343 - Undeclared: declare:from: cleanup, first step #15290 https://github.com/pharo-project/pharo/pull/15290 - Remove private system announcer #15355 https://github.com/pharo-project/pharo/pull/15355 - do not hardcode global environment in #removeFromSystem: #15357 https://github.com/pharo-project/pharo/pull/15357 - Add binary selector alias for #newFromArray: for Collection #15316 https://github.com/pharo-project/pharo/pull/15316 # Traits - Traits: Fix bug in dictionary rebuild #15353 https://github.com/pharo-project/pharo/pull/15353 - Speedup #explicitRequirement #15322 https://github.com/pharo-project/pharo/pull/15322 - Traits: Speedup #isFromTrait #15318 https://github.com/pharo-project/pharo/pull/15318 - Rename TraitsV2 into Traits #15319 https://github.com/pharo-project/pharo/pull/15319 - Traits: Do a new cleaning pass over the traits tests and fix file out bug #15304 https://github.com/pharo-project/pharo/pull/15304 - Remove TraitsV2-Compatibility #15325 https://github.com/pharo-project/pharo/pull/15325 # Refactoring Engine - Migrate rename class to driver #15394 https://github.com/pharo-project/pharo/pull/15394 - remove deadcode from driver package #15390 https://github.com/pharo-project/pharo/pull/15390 - RB: A simple fix for abstract methods. No brainer. #15331 https://github.com/pharo-project/pharo/pull/15331 - Remove unused SycRemoveMethodCommand #15337 https://github.com/pharo-project/pharo/pull/15337 - Refactorings: Follow package name convention #15340 https://github.com/pharo-project/pharo/pull/15340 # Tests - Fix #testUndeclared in case of failure #15356 https://github.com/pharo-project/pharo/pull/15356 - Release tests: Improve testNoSentButNotImplementedSelectors #15341 https://github.com/pharo-project/pharo/pull/15341 - Use a test resource for SystemDependenciesTest #15308 https://github.com/pharo-project/pharo/pull/15308 # Cleanup - Some cleanup of CodeExport #15380 https://github.com/pharo-project/pharo/pull/15380 - Deprecate gofer command line #15347 https://github.com/pharo-project/pharo/pull/15347 - Clean AST-Core #15358 https://github.com/pharo-project/pharo/pull/15358 - Remove #isUnitTest #15349 https://github.com/pharo-project/pharo/pull/15349 - Manifest builder should not rely on categories #15324 https://github.com/pharo-project/pharo/pull/15324 - Remove workspace #15284 https://github.com/pharo-project/pharo/pull/15284 - Deprecate TRBProgramNodeVisitor #15305 https://github.com/pharo-project/pharo/pull/15305 - Refactor #inspectionIcons on ThemeIcons to use the #icons accessor #609 https://github.com/pharo-spec/NewTools/pull/609 # Cleanup Trivail - fix icons #15371 https://github.com/pharo-project/pharo/pull/15371 - use allButFirst and simplify expression #15382 https://github.com/pharo-project/pharo/pull/15382 - use allButFirst instead of copyFrom:to: #15369 https://github.com/pharo-project/pharo/pull/15369 - use allButFirst instead of copyFrom:to: #15366 https://github.com/pharo-project/pharo/pull/15366 - Remove deprecated calls #15354 https://github.com/pharo-project/pharo/pull/15354 - Cleanup: usage of notNil in Text-Scanning, Rubric, HeuristicCompletion, NECCompletion, Growl, Regex #15296 https://github.com/pharo-project/pharo/pull/15296 - Cleanup: usage of notNil in Calypse, Microdown, System-Changes, STON, System-Settings #15297 https://github.com/pharo-project/pharo/pull/15297 - Cleanup: Usage of "notNil" and "isEmpty not" in Zinc / Zodiac #15253 https://github.com/pharo-project/pharo/pull/15253

Nov. 17, 2023

Re: [Pharo-vm] Stack overflow support?

by stephane ducasse

Thanks. We will digest and discuss internally. S > On 14 Nov 2023, at 07:19, Daniel Slomovits <daniels220(a)gmail.com> wrote: > > Hello Stephane, > > A simple example like thatâreally any case caused by common types of mistake, even if the method involved is more complexâwouldn't get to run for a couple seconds. It would error out after a small fraction of a second, since the interrupt is issued immediately upon exceeding the limit (which by default isâoops, 64k slots, 256kB, at most ~10k stack frames, but realistically less due to args and temps), and it just doesn't take that long to execute most recursive methods 10k levels deep. > > At this point it would raise the limit a little for that process and signal a StackOverflow exception. In production this likely would mean killing the process but that's up to application code. In development you get a walkback like any other exception. Because the stack is kept to a relatively reasonable size, I've never had debugger performance be an issueâit's noticeably more sluggish, but we're talking about taking 200-300ms to respond instead of imperceptible. Similarly for GC pressureâeven if you have dozens of processes all maxed out on stack, this would still be less than the memory used to start a base image. > > Certainly there are scenarios that can cause problemsâan error while printing an object will stop you from opening a debugger, yes. However a StackOverflow in particular isn't any worse than any other exceptionâin all cases you'll get a walkback, hit "Debug", and get another walkback instead of a debugger. If you hit "Terminate" at that point, it might leave the original process in a zombie state and/or a hidden debugger window, but you can kill it from the Process Monitor/close it from the Window menu and be fine. And even this is because Dolphin doesn't safeguard printing in the development tools the way Pharo doesâif Pharo adopted stack overflow handling like Dolphin's, a stack overflow likely wouldn't even stop you from opening a debugger, it would just make it a little slow, then something would show up with "error printing: a StackOverflow" or the like. > > Hope that helps. > > Daniel > > On Sun, Nov 12, 2023 at 4:53â¯AM stephane ducasse <stephane.ducasse(a)inria.fr <mailto:stephane.ducasse@inria.fr>> wrote: >> Hi daniel >> >> Thanks for the feedback. >> May be you wrote it but I could not really understand. >> >> How dolphin handled >> >>>> ``` >>>> A >> foo >>>> ^ self foo >>>> ``` >> >> That is let to run a couple of seconds? >> Did they kill the process? >> >> In Pharo we do have an interrupt but >> >>>> But it could happen that, >>>> - the stack is so big that the debugger is very sluggish (best-case scenario) >>>> - the VM is just flooded doing GCs so maybe the Ctrl dot event does not even arrive at Pharo or the trigger >>>> - if the recursion is hit when printing an object (which is more common than you could imagine), opening the debugger could trigger a new recursion and never give back the control to the user >> >> >> S >> >>> On 11 Nov 2023, at 05:10, Daniel Slomovits <daniels220(a)gmail.com <mailto:daniels220@gmail.com>> wrote: >>> >>> I think this is a great idea. I've mostly used Dolphin Smalltalk, which is actually a strict stack machine under the hood (it has a context-like introspection API but the stack is explicitly the canonical form), so it's more-or-less forced to implement a limit of some kind. When I started using Pharo I triggered a couple stack overflows by mistake, and was frustrated by the fact that at first what happened was...nothing, everything seemed fine, my code just didn't work. And then half a minute later Pharo gets extremely slow and I notice it's using 2GB of memory and by then it's too late and I have to kill the image. Getting a more-or-less immediate error would be much more user-friendly IMO. >>> >>> A couple things to learn from Dolphin's implementation, I think: >>> When a stack overflow is detected, the resulting interrupt raises that process' stack limit by a significant amount (though by less than the original limitâIOW it doesn't double, but it's not just a couple more frames either) before signaling the exception, precisely so that exception handling can occur without triggering another stack-overflow event. A further refinement could be that if a second stack overflow is detected, we directly invoke more basic recoveryâthis could mean an emergency evaluator, terminating the offending process and opening a post-mortem with a textual stack dump (ugh! but at least it's predictable), etc. >>> I don't think we should worry too much about refining what exactly the limit is. 10x as much stack as 99% of code will ever use, is still a tiny amount compared to consuming all available memory with Contexts. At least, if I'm understanding the graph/data correctly. That's 36kB of stack space, right? Not 36k frames/contexts deep? With each context being six slots plus args/temps, 36kB is 500-750 frames on a 64-bit VM (in stack representationâcontexts add object-header overhead but we don't reify them unless we have to). For reference, Dolphin's limit is 64kB, but that's a 32-bit VM, so the equivalent for 64-bit would be 128kB...but because Pharo can spill contexts to the stack, the limit could easily be 1MB, or a fixed number of frames designed to approximate thatâstill a tiny amount of memory overall, and still will be hit near-instantly by true infinite recursion, but lots of breathing room for most use cases. >>> Actually, this did get me to thinking...the stack depth of a Pharo process is not necessarily easy/cheap to compute in the general case, without caching a lot of information on intermediate contexts. In most cases the context chain acts as a proper stack, but not alwaysâmethods like Process>>on:do:, and some even more esoteric ones I forget off the top of my head, make modifications far away from the top context and may splice context chains together in odd ways. Perhaps a more flexible limit would be betterâone that is triggered by allocating more than a certain number of contexts total, and examines running Processes in detail to find the culprit at that point. >>> >>> On Thu, Nov 9, 2023 at 4:38â¯AM Guillermo Polito <guillermopolito(a)gmail.com <mailto:guillermopolito@gmail.com>> wrote: >>>> Hi all, >>>> >>>> We started (with many interruptions over the last months) working a bit with Stephane on understanding what is the (positive and negative) impact of stack-overflow support in Pharo. >>>> The key idea is that if a process consumes too much stack (potentially because of an infinite recursion) then the process should stop with an exception. >>>> >>>> ## Why we want better stack consumption control >>>> >>>> This idea comes up to solve issues that are pretty common and hit especially newbies. >>>> For example, imagine you accidentally write an accessor such as >>>> >>>> ``` >>>> A >> foo >>>> ^ self foo >>>> ``` >>>> >>>> Students do this all the time, and Iâve also seen it in experienced people who go too fast :). >>>> More importantly, such recursions could happen also with not-so-obvious indirect recursions (a sends b, b sends c, c sends a), and these could hit anybody. >>>> >>>> This is aggravated because the current execution model allows us to have infinite stacks âmeaning: limited by available memory only. >>>> This is indeed a nice feature for many use cases but it has its own drawbacks when one of these kind of recursions are hit: >>>> - code just loops forever taking space in the stack >>>> - when there is no more stack space, context objects are created and moved to the heap >>>> - but those contexts are strongly held, so they are never GCed and take up extra space >>>> - even worse! they are there adding more work to the GC every time and making the GC run more often looking for space that is not there >>>> >>>> ## Why Ctrl-dot does not always work >>>> >>>> Of course, super users know there is this âCtrl dotâ hidden feature that should help you recover from this. >>>> First, let's take out of the equation that this is only known by super users. >>>> Now, in this situation, when Ctrl-dot is hit it will trigger a handler that suspends the problematic process and opens a debugger on it. >>>> But it could happen that, >>>> - the stack is so big that the debugger is very sluggish (best-case scenario) >>>> - the VM is just flooded doing GCs so maybe the Ctrl dot event does not even arrive at Pharo or the trigger >>>> - if the recursion is hit when printing an object (which is more common than you could imagine), opening the debugger could trigger a new recursion and never give back the control to the user >>>> >>>> ## What are we working on >>>> >>>> The main idea here is: Can we have a simple and efficient way to prevent such kinds of situations? >>>> >>>> After many discussions around detecting recursion, we kinda arrived at the simple solution of just detecting a stack overflow. >>>> The solution is easy to understand (because itâs like other languages work) and easy to implement because there is already support for that. >>>> But this leaves open two questions: >>>> - what happens when people want to use the âinfinite stackâ feature? >>>> - when should a process stack overflow? What is a sensitive default value? >>>> >>>> Our draft implementation here https://github.com/pharo-project/pharo-vm/pull/710 does the following to cope with this: >>>> - we can now parametrize the size of the stack (of each stack page to be more accurate) when the VM starts up >>>> - the stack overflow check can be disabled per process >>>> >>>> We also are running experiments to see what could be a sensitive stack size for our normal usages. Here, for example, we ran almost all test cases in Pharo separately (one suite per line below), and we observed how many tests broke (x-axis) with different stack sizes (y-axis). >>>> Here we see that most test suites require at least 20-24k to run properly, some go up to 36k of stack before converging (i.e., the number of broken tests does not change). >>>> >>>> <ImagenPegada-10.tiff> >>>> >>>> Youâll notice in the graph that There are some scenarios that break all the time. This is because exception handling itself is recursive and may produce more stack overflows depending on the size of the stack between the exception and the exception handler. >>>> So some more work is still required, mostly changing Pharo libraries to properly support this. For example: >>>> - should tests run in a fresh process with a fresh stack? >>>> - should the exception mechanism use less recursion? >>>> - resumable exceptions add stack pressure because they do not âunstackâ until the exception is finally handled, meaning that the stack used by exception handling just adds up to the stack of the original code, can we do better here? >>>> >>>> Probably there are more interesting questions here, thatâs the âwhy" behind this email. >>>> Iâm interested in opinions and scenarios you may come up with that should be taken into account. >>>> >>>> Cheers, >>>> Guille >>> _______________________________________________ >>> Pharo-vm mailing list -- pharo-vm(a)lists.pharo.org <mailto:pharo-vm@lists.pharo.org> >>> To unsubscribe send an email to pharo-vm-leave(a)lists.pharo.org <mailto:pharo-vm-leave@lists.pharo.org> >> >> -------------------------------------------- >> StÃ©phane Ducasse >> http://stephane.ducasse.free.fr <http://stephane.ducasse.free.fr/> / http://www.pharo.org <http://www.pharo.org/> >> 03 59 35 87 52 >> Assistant: Aurore Dalle >> FAX 03 59 57 78 50 >> TEL 03 59 35 86 16 >> S. Ducasse - Inria >> 40, avenue Halley, >> Parc Scientifique de la Haute Borne, BÃ¢t.A, Park Plaza >> Villeneuve d'Ascq 59650 >> France >> >> >> > _______________________________________________ > Pharo-vm mailing list -- pharo-vm(a)lists.pharo.org > To unsubscribe send an email to pharo-vm-leave(a)lists.pharo.org -------------------------------------------- StÃ©phane Ducasse http://stephane.ducasse.free.fr / http://www.pharo.org 03 59 35 87 52 Assistant: Aurore Dalle FAX 03 59 57 78 50 TEL 03 59 35 86 16 S. Ducasse - Inria 40, avenue Halley, Parc Scientifique de la Haute Borne, BÃ¢t.A, Park Plaza Villeneuve d'Ascq 59650 France

Nov. 15, 2023

Re: New blog post on all:objects all: theTime

by Noury Bouraqadi

Thank you Koen for update. Keep pushing. Noury On Nov 15 2023, at 8:25 am, Koen De Hondt <koen(a)all-objects-all-the-time.st> wrote: > Dear Pharo users and developers, > > After my presentation at ESUGâ23, which has been added to my blog (https://all-objects-all-the-time.st/#/blog/posts/4) too, I wrote an evaluation of Calypso (https://all-objects-all-the-time.st/#/blog/posts/5) in which I refer to results of the Pharo Browser Survey that I presented at ESUGâ23. > > It is a long post. It will be the basis for more posts on browsers in Pharo. > > My blog does not support commenting. If you have any feedback or comment, please feel free to send me a message. > > Happy reading! > > Ciao, > Koen

Nov. 15, 2023

New blog post on all:objects all: theTime

by Koen De Hondt

Dear Pharo users and developers, After my presentation at ESUGâ23, which has been added to my blog <https://all-objects-all-the-time.st/#/blog/posts/4> too, I wrote an evaluation of Calypso <https://all-objects-all-the-time.st/#/blog/posts/5>, in which I refer to results of the Pharo Browser Survey that I presented at ESUGâ23. It is a long post. It will be the basis for more posts on browsers in Pharo. My blog does not support commenting. If you have any feedback or comment, please feel free to send me a message. Happy reading! Ciao, Koen

Nov. 15, 2023

Re: [Pharo-vm] Stack overflow support?

by Daniel Slomovits

Hello Stephane, A simple example like thatâreally any case caused by common types of mistake, even if the method involved is more complexâwouldn't get to run for a couple seconds. It would error out after a small fraction of a second, since the interrupt is issued immediately upon exceeding the limit (which by default isâoops, 64k* slots*, 256kB, at most ~10k stack frames, but realistically less due to args and temps), and it just doesn't take that long to execute most recursive methods 10k levels deep. At this point it would raise the limit a little for that process and signal a StackOverflow exception. In production this likely would mean killing the process but that's up to application code. In development you get a walkback like any other exception. Because the stack is kept to a *relatively* reasonable size, I've never had debugger performance be an issueâit's noticeably more sluggish, but we're talking about taking 200-300ms to respond instead of imperceptible. Similarly for GC pressureâeven if you have dozens of processes all maxed out on stack, this would still be less than the memory used to start a base image. Certainly there are scenarios that can cause problemsâan error while printing an object will stop you from opening a debugger, yes. However a StackOverflow in particular isn't any worse than any other exceptionâin all cases you'll get a walkback, hit "Debug", and get another walkback instead of a debugger. If you hit "Terminate" at that point, it might leave the original process in a zombie state and/or a hidden debugger window, but you can kill it from the Process Monitor/close it from the Window menu and be fine. And even this is because Dolphin doesn't safeguard printing in the development tools the way Pharo doesâif Pharo adopted stack overflow handling like Dolphin's, a stack overflow likely wouldn't even stop you from opening a debugger, it would just make it a *little* slow, then something would show up with "error printing: a StackOverflow" or the like. Hope that helps. Daniel On Sun, Nov 12, 2023 at 4:53â¯AM stephane ducasse <stephane.ducasse(a)inria.fr> wrote: > Hi daniel > > Thanks for the feedback. > May be you wrote it but I could not really understand. > > How dolphin handled > > ``` > A >> foo > ^ self foo > ``` > > > That is let to run a couple of seconds? > Did they kill the process? > > In Pharo we do have an interrupt but > > But it could happen that, > - the stack is so big that the debugger is very sluggish (best-case > scenario) > - the VM is just flooded doing GCs so maybe the Ctrl dot event does not > even arrive at Pharo or the trigger > - if the recursion is hit when printing an object (which is more common > than you could imagine), opening the debugger could trigger a new recursion > and never give back the control to the user > > > S > > On 11 Nov 2023, at 05:10, Daniel Slomovits <daniels220(a)gmail.com> wrote: > > I think this is a great idea. I've mostly used Dolphin Smalltalk, which is > actually a strict stack machine under the hood (it has a context-like > introspection API but the stack is explicitly the canonical form), so it's > more-or-less forced to implement a limit of some kind. When I started using > Pharo I triggered a couple stack overflows by mistake, and was frustrated > by the fact that at first what happened was...nothing, everything seemed > fine, my code just didn't work. And then half a minute later Pharo gets > extremely slow and I notice it's using 2GB of memory and by then it's too > late and I have to kill the image. Getting a more-or-less immediate error > would be much more user-friendly IMO. > > A couple things to learn from Dolphin's implementation, I think: > > 1. When a stack overflow is detected, the resulting interrupt raises > that process' stack limit by a significant amount (though by less than the > original limitâIOW it doesn't double, but it's not just a couple more > frames either) before signaling the exception, precisely so that exception > handling can occur without triggering another stack-overflow event. A > further refinement could be that if a second stack overflow *is* detected, > we directly invoke more basic recoveryâthis could mean an emergency > evaluator, terminating the offending process and opening a post-mortem with > a textual stack dump (ugh! but at least it's predictable), etc. > 2. I don't think we should worry too much about refining what exactly > the limit is. 10x as much stack as 99% of code will ever use, is still a > tiny amount compared to consuming all available memory with Contexts. At > least, if I'm understanding the graph/data correctly. That's 36kB of stack > space, right? Not 36k frames/contexts deep? With each context being six > slots plus args/temps, 36kB is 500-750 frames on a 64-bit VM (in stack > representationâcontexts add object-header overhead but we don't reify them > unless we have to). For reference, Dolphin's limit is 64kB, but that's a > 32-bit VM, so the equivalent for 64-bit would be 128kB...but because Pharo > can spill contexts to the stack, the limit could easily be 1MB, or a fixed > number of frames designed to approximate thatâstill a tiny amount of memory > overall, and still will be hit near-instantly by true infinite recursion, > but lots of breathing room for most use cases. > > Actually, this did get me to thinking...the stack depth of a Pharo process > is not necessarily easy/cheap to compute in the general case, without > caching a lot of information on intermediate contexts. In most cases the > context chain acts as a proper stack, but *not always*âmethods like > Process>>on:do:, and some even more esoteric ones I forget off the top of > my head, make modifications far away from the top context and may splice > context chains together in odd ways. Perhaps a more flexible limit would be > betterâone that is triggered by allocating more than a certain number of > contexts *total*, and examines running Processes in detail to find the > culprit at that point. > > On Thu, Nov 9, 2023 at 4:38â¯AM Guillermo Polito <guillermopolito(a)gmail.com> > wrote: > >> Hi all, >> >> We started (with many interruptions over the last months) working a bit >> with Stephane on understanding what is the (positive and negative) impact >> of stack-overflow support in Pharo. >> The key idea is that if a process consumes too much stack (potentially >> because of an infinite recursion) then the process should stop with an >> exception. >> >> ## Why we want better stack consumption control >> >> This idea comes up to solve issues that are pretty common and hit >> especially newbies. >> For example, imagine you accidentally write an accessor such as >> >> ``` >> A >> foo >> ^ self foo >> ``` >> >> Students do this all the time, and Iâve also seen it in experienced >> people who go too fast :). >> More importantly, such recursions could happen also with not-so-obvious >> indirect recursions (a sends b, b sends c, c sends a), and these could hit >> anybody. >> >> This is aggravated because the current execution model allows us to have >> infinite stacks âmeaning: limited by available memory only. >> This is indeed a nice feature for many use cases but it has its own >> drawbacks when one of these kind of recursions are hit: >> - code just loops forever taking space in the stack >> - when there is no more stack space, context objects are created and >> moved to the heap >> - but those contexts are strongly held, so they are never GCed and take >> up extra space >> - even worse! they are there adding more work to the GC every time and >> making the GC run more often looking for space that is not there >> >> ## Why Ctrl-dot does not always work >> >> Of course, super users know there is this âCtrl dotâ hidden feature that >> should help you recover from this. >> First, let's take out of the equation that this is only known by super >> users. >> Now, in this situation, when Ctrl-dot is hit it will trigger a handler >> that suspends the problematic process and opens a debugger on it. >> But it could happen that, >> - the stack is so big that the debugger is very sluggish (best-case >> scenario) >> - the VM is just flooded doing GCs so maybe the Ctrl dot event does not >> even arrive at Pharo or the trigger >> - if the recursion is hit when printing an object (which is more common >> than you could imagine), opening the debugger could trigger a new recursion >> and never give back the control to the user >> >> ## What are we working on >> >> The main idea here is: Can we have a simple and efficient way to prevent >> such kinds of situations? >> >> After many discussions around detecting recursion, we kinda arrived at >> the simple solution of just detecting a stack overflow. >> The solution is easy to understand (because itâs like other languages >> work) and easy to implement because there is already support for that. >> But this leaves open two questions: >> - what happens when people want to use the âinfinite stackâ feature? >> - when should a process stack overflow? What is a sensitive default >> value? >> >> Our draft implementation here >> https://github.com/pharo-project/pharo-vm/pull/710 does the following to >> cope with this: >> - we can now parametrize the size of the stack (of each stack page to be >> more accurate) when the VM starts up >> - the stack overflow check can be disabled per process >> >> We also are running experiments to see what could be a sensitive stack >> size for our normal usages. Here, for example, we ran almost all test cases >> in Pharo separately (one suite per line below), and we observed how many >> tests broke (x-axis) with different stack sizes (y-axis). >> Here we see that most test suites require at least 20-24k to run >> properly, some go up to 36k of stack before converging (i.e., the number of >> broken tests does not change). >> >> <ImagenPegada-10.tiff> >> >> Youâll notice in the graph that There are some scenarios that break all >> the time. This is because exception handling itself is recursive and may >> produce more stack overflows depending on the size of the stack between the >> exception and the exception handler. >> So some more work is still required, mostly changing Pharo libraries to >> properly support this. For example: >> - should tests run in a fresh process with a fresh stack? >> - should the exception mechanism use less recursion? >> - resumable exceptions add stack pressure because they do not âunstackâ >> until the exception is finally handled, meaning that the stack used by >> exception handling just adds up to the stack of the original code, can we >> do better here? >> >> Probably there are more interesting questions here, thatâs the âwhy" >> behind this email. >> Iâm interested in opinions and scenarios you may come up with that should >> be taken into account. >> >> Cheers, >> Guille >> > _______________________________________________ > Pharo-vm mailing list -- pharo-vm(a)lists.pharo.org > To unsubscribe send an email to pharo-vm-leave(a)lists.pharo.org > > > -------------------------------------------- > StÃ©phane Ducasse > http://stephane.ducasse.free.fr / http://www.pharo.org > 03 59 35 87 52 > Assistant: Aurore Dalle > FAX 03 59 57 78 50 > TEL 03 59 35 86 16 > S. Ducasse - Inria > 40, avenue Halley, > Parc Scientifique de la Haute Borne, BÃ¢t.A, Park Plaza > Villeneuve d'Ascq 59650 > France > > > >

Nov. 14, 2023

Re: [Pharo-vm] Stack overflow support?

by stephane ducasse

Hi daniel Thanks for the feedback. May be you wrote it but I could not really understand. How dolphin handled >> ``` >> A >> foo >> ^ self foo >> ``` That is let to run a couple of seconds? Did they kill the process? In Pharo we do have an interrupt but >> But it could happen that, >> - the stack is so big that the debugger is very sluggish (best-case scenario) >> - the VM is just flooded doing GCs so maybe the Ctrl dot event does not even arrive at Pharo or the trigger >> - if the recursion is hit when printing an object (which is more common than you could imagine), opening the debugger could trigger a new recursion and never give back the control to the user S > On 11 Nov 2023, at 05:10, Daniel Slomovits <daniels220(a)gmail.com> wrote: > > I think this is a great idea. I've mostly used Dolphin Smalltalk, which is actually a strict stack machine under the hood (it has a context-like introspection API but the stack is explicitly the canonical form), so it's more-or-less forced to implement a limit of some kind. When I started using Pharo I triggered a couple stack overflows by mistake, and was frustrated by the fact that at first what happened was...nothing, everything seemed fine, my code just didn't work. And then half a minute later Pharo gets extremely slow and I notice it's using 2GB of memory and by then it's too late and I have to kill the image. Getting a more-or-less immediate error would be much more user-friendly IMO. > > A couple things to learn from Dolphin's implementation, I think: > When a stack overflow is detected, the resulting interrupt raises that process' stack limit by a significant amount (though by less than the original limitâIOW it doesn't double, but it's not just a couple more frames either) before signaling the exception, precisely so that exception handling can occur without triggering another stack-overflow event. A further refinement could be that if a second stack overflow is detected, we directly invoke more basic recoveryâthis could mean an emergency evaluator, terminating the offending process and opening a post-mortem with a textual stack dump (ugh! but at least it's predictable), etc. > I don't think we should worry too much about refining what exactly the limit is. 10x as much stack as 99% of code will ever use, is still a tiny amount compared to consuming all available memory with Contexts. At least, if I'm understanding the graph/data correctly. That's 36kB of stack space, right? Not 36k frames/contexts deep? With each context being six slots plus args/temps, 36kB is 500-750 frames on a 64-bit VM (in stack representationâcontexts add object-header overhead but we don't reify them unless we have to). For reference, Dolphin's limit is 64kB, but that's a 32-bit VM, so the equivalent for 64-bit would be 128kB...but because Pharo can spill contexts to the stack, the limit could easily be 1MB, or a fixed number of frames designed to approximate thatâstill a tiny amount of memory overall, and still will be hit near-instantly by true infinite recursion, but lots of breathing room for most use cases. > Actually, this did get me to thinking...the stack depth of a Pharo process is not necessarily easy/cheap to compute in the general case, without caching a lot of information on intermediate contexts. In most cases the context chain acts as a proper stack, but not alwaysâmethods like Process>>on:do:, and some even more esoteric ones I forget off the top of my head, make modifications far away from the top context and may splice context chains together in odd ways. Perhaps a more flexible limit would be betterâone that is triggered by allocating more than a certain number of contexts total, and examines running Processes in detail to find the culprit at that point. > > On Thu, Nov 9, 2023 at 4:38â¯AM Guillermo Polito <guillermopolito(a)gmail.com <mailto:guillermopolito@gmail.com>> wrote: >> Hi all, >> >> We started (with many interruptions over the last months) working a bit with Stephane on understanding what is the (positive and negative) impact of stack-overflow support in Pharo. >> The key idea is that if a process consumes too much stack (potentially because of an infinite recursion) then the process should stop with an exception. >> >> ## Why we want better stack consumption control >> >> This idea comes up to solve issues that are pretty common and hit especially newbies. >> For example, imagine you accidentally write an accessor such as >> >> ``` >> A >> foo >> ^ self foo >> ``` >> >> Students do this all the time, and Iâve also seen it in experienced people who go too fast :). >> More importantly, such recursions could happen also with not-so-obvious indirect recursions (a sends b, b sends c, c sends a), and these could hit anybody. >> >> This is aggravated because the current execution model allows us to have infinite stacks âmeaning: limited by available memory only. >> This is indeed a nice feature for many use cases but it has its own drawbacks when one of these kind of recursions are hit: >> - code just loops forever taking space in the stack >> - when there is no more stack space, context objects are created and moved to the heap >> - but those contexts are strongly held, so they are never GCed and take up extra space >> - even worse! they are there adding more work to the GC every time and making the GC run more often looking for space that is not there >> >> ## Why Ctrl-dot does not always work >> >> Of course, super users know there is this âCtrl dotâ hidden feature that should help you recover from this. >> First, let's take out of the equation that this is only known by super users. >> Now, in this situation, when Ctrl-dot is hit it will trigger a handler that suspends the problematic process and opens a debugger on it. >> But it could happen that, >> - the stack is so big that the debugger is very sluggish (best-case scenario) >> - the VM is just flooded doing GCs so maybe the Ctrl dot event does not even arrive at Pharo or the trigger >> - if the recursion is hit when printing an object (which is more common than you could imagine), opening the debugger could trigger a new recursion and never give back the control to the user >> >> ## What are we working on >> >> The main idea here is: Can we have a simple and efficient way to prevent such kinds of situations? >> >> After many discussions around detecting recursion, we kinda arrived at the simple solution of just detecting a stack overflow. >> The solution is easy to understand (because itâs like other languages work) and easy to implement because there is already support for that. >> But this leaves open two questions: >> - what happens when people want to use the âinfinite stackâ feature? >> - when should a process stack overflow? What is a sensitive default value? >> >> Our draft implementation here https://github.com/pharo-project/pharo-vm/pull/710 does the following to cope with this: >> - we can now parametrize the size of the stack (of each stack page to be more accurate) when the VM starts up >> - the stack overflow check can be disabled per process >> >> We also are running experiments to see what could be a sensitive stack size for our normal usages. Here, for example, we ran almost all test cases in Pharo separately (one suite per line below), and we observed how many tests broke (x-axis) with different stack sizes (y-axis). >> Here we see that most test suites require at least 20-24k to run properly, some go up to 36k of stack before converging (i.e., the number of broken tests does not change). >> >> <ImagenPegada-10.tiff> >> >> Youâll notice in the graph that There are some scenarios that break all the time. This is because exception handling itself is recursive and may produce more stack overflows depending on the size of the stack between the exception and the exception handler. >> So some more work is still required, mostly changing Pharo libraries to properly support this. For example: >> - should tests run in a fresh process with a fresh stack? >> - should the exception mechanism use less recursion? >> - resumable exceptions add stack pressure because they do not âunstackâ until the exception is finally handled, meaning that the stack used by exception handling just adds up to the stack of the original code, can we do better here? >> >> Probably there are more interesting questions here, thatâs the âwhy" behind this email. >> Iâm interested in opinions and scenarios you may come up with that should be taken into account. >> >> Cheers, >> Guille > _______________________________________________ > Pharo-vm mailing list -- pharo-vm(a)lists.pharo.org > To unsubscribe send an email to pharo-vm-leave(a)lists.pharo.org -------------------------------------------- StÃ©phane Ducasse http://stephane.ducasse.free.fr / http://www.pharo.org 03 59 35 87 52 Assistant: Aurore Dalle FAX 03 59 57 78 50 TEL 03 59 35 86 16 S. Ducasse - Inria 40, avenue Halley, Parc Scientifique de la Haute Borne, BÃ¢t.A, Park Plaza Villeneuve d'Ascq 59650 France

Nov. 12, 2023

Re: Stack overflow support?

by Daniel Slomovits

I think this is a great idea. I've mostly used Dolphin Smalltalk, which is actually a strict stack machine under the hood (it has a context-like introspection API but the stack is explicitly the canonical form), so it's more-or-less forced to implement a limit of some kind. When I started using Pharo I triggered a couple stack overflows by mistake, and was frustrated by the fact that at first what happened was...nothing, everything seemed fine, my code just didn't work. And then half a minute later Pharo gets extremely slow and I notice it's using 2GB of memory and by then it's too late and I have to kill the image. Getting a more-or-less immediate error would be much more user-friendly IMO. A couple things to learn from Dolphin's implementation, I think: 1. When a stack overflow is detected, the resulting interrupt raises that process' stack limit by a significant amount (though by less than the original limitâIOW it doesn't double, but it's not just a couple more frames either) before signaling the exception, precisely so that exception handling can occur without triggering another stack-overflow event. A further refinement could be that if a second stack overflow *is* detected, we directly invoke more basic recoveryâthis could mean an emergency evaluator, terminating the offending process and opening a post-mortem with a textual stack dump (ugh! but at least it's predictable), etc. 2. I don't think we should worry too much about refining what exactly the limit is. 10x as much stack as 99% of code will ever use, is still a tiny amount compared to consuming all available memory with Contexts. At least, if I'm understanding the graph/data correctly. That's 36kB of stack space, right? Not 36k frames/contexts deep? With each context being six slots plus args/temps, 36kB is 500-750 frames on a 64-bit VM (in stack representationâcontexts add object-header overhead but we don't reify them unless we have to). For reference, Dolphin's limit is 64kB, but that's a 32-bit VM, so the equivalent for 64-bit would be 128kB...but because Pharo can spill contexts to the stack, the limit could easily be 1MB, or a fixed number of frames designed to approximate thatâstill a tiny amount of memory overall, and still will be hit near-instantly by true infinite recursion, but lots of breathing room for most use cases. Actually, this did get me to thinking...the stack depth of a Pharo process is not necessarily easy/cheap to compute in the general case, without caching a lot of information on intermediate contexts. In most cases the context chain acts as a proper stack, but *not always*âmethods like Process>>on:do:, and some even more esoteric ones I forget off the top of my head, make modifications far away from the top context and may splice context chains together in odd ways. Perhaps a more flexible limit would be betterâone that is triggered by allocating more than a certain number of contexts *total*, and examines running Processes in detail to find the culprit at that point. On Thu, Nov 9, 2023 at 4:38â¯AM Guillermo Polito <guillermopolito(a)gmail.com> wrote: > Hi all, > > We started (with many interruptions over the last months) working a bit > with Stephane on understanding what is the (positive and negative) impact > of stack-overflow support in Pharo. > The key idea is that if a process consumes too much stack (potentially > because of an infinite recursion) then the process should stop with an > exception. > > ## Why we want better stack consumption control > > This idea comes up to solve issues that are pretty common and hit > especially newbies. > For example, imagine you accidentally write an accessor such as > > ``` > A >> foo > ^ self foo > ``` > > Students do this all the time, and Iâve also seen it in experienced people > who go too fast :). > More importantly, such recursions could happen also with not-so-obvious > indirect recursions (a sends b, b sends c, c sends a), and these could hit > anybody. > > This is aggravated because the current execution model allows us to have > infinite stacks âmeaning: limited by available memory only. > This is indeed a nice feature for many use cases but it has its own > drawbacks when one of these kind of recursions are hit: > - code just loops forever taking space in the stack > - when there is no more stack space, context objects are created and > moved to the heap > - but those contexts are strongly held, so they are never GCed and take > up extra space > - even worse! they are there adding more work to the GC every time and > making the GC run more often looking for space that is not there > > ## Why Ctrl-dot does not always work > > Of course, super users know there is this âCtrl dotâ hidden feature that > should help you recover from this. > First, let's take out of the equation that this is only known by super > users. > Now, in this situation, when Ctrl-dot is hit it will trigger a handler > that suspends the problematic process and opens a debugger on it. > But it could happen that, > - the stack is so big that the debugger is very sluggish (best-case > scenario) > - the VM is just flooded doing GCs so maybe the Ctrl dot event does not > even arrive at Pharo or the trigger > - if the recursion is hit when printing an object (which is more common > than you could imagine), opening the debugger could trigger a new recursion > and never give back the control to the user > > ## What are we working on > > The main idea here is: Can we have a simple and efficient way to prevent > such kinds of situations? > > After many discussions around detecting recursion, we kinda arrived at the > simple solution of just detecting a stack overflow. > The solution is easy to understand (because itâs like other languages > work) and easy to implement because there is already support for that. > But this leaves open two questions: > - what happens when people want to use the âinfinite stackâ feature? > - when should a process stack overflow? What is a sensitive default value? > > Our draft implementation here > https://github.com/pharo-project/pharo-vm/pull/710 does the following to > cope with this: > - we can now parametrize the size of the stack (of each stack page to be > more accurate) when the VM starts up > - the stack overflow check can be disabled per process > > We also are running experiments to see what could be a sensitive stack > size for our normal usages. Here, for example, we ran almost all test cases > in Pharo separately (one suite per line below), and we observed how many > tests broke (x-axis) with different stack sizes (y-axis). > Here we see that most test suites require at least 20-24k to run properly, > some go up to 36k of stack before converging (i.e., the number of broken > tests does not change). > > > Youâll notice in the graph that There are some scenarios that break all > the time. This is because exception handling itself is recursive and may > produce more stack overflows depending on the size of the stack between the > exception and the exception handler. > So some more work is still required, mostly changing Pharo libraries to > properly support this. For example: > - should tests run in a fresh process with a fresh stack? > - should the exception mechanism use less recursion? > - resumable exceptions add stack pressure because they do not âunstackâ > until the exception is finally handled, meaning that the stack used by > exception handling just adds up to the stack of the original code, can we > do better here? > > Probably there are more interesting questions here, thatâs the âwhy" > behind this email. > Iâm interested in opinions and scenarios you may come up with that should > be taken into account. > > Cheers, > Guille >

Nov. 11, 2023

This week (45/2023) on the Pharo Issue Tracker

by Marcus Denker

# Feature - Proposal: new message browser layout #15229 https://github.com/pharo-project/pharo/pull/15229 # Fixes - Calypso: Fix refresh of project view #15281 https://github.com/pharo-project/pharo/pull/15281 - Stay in calypso project view when selecting a class #15247 https://github.com/pharo-project/pharo/pull/15247 - Shout: Fix styling of symbols representing a deprecated global variable #15274 https://github.com/pharo-project/pharo/pull/15274 - Calypso: Add Undefined project #15246 https://github.com/pharo-project/pharo/pull/15246 - Fix clean blocks with dead context on debugger #15153 https://github.com/pharo-project/pharo/pull/15153 - Fix âspanâ in BalloonEngine not being wider than the âdestFormâ #15236 https://github.com/pharo-project/pharo/pull/15236 - Calypso: Fix taskbar label #15226 https://github.com/pharo-project/pharo/pull/15226 # Refactoring Engine - Feat: Add Override method transformation #15225 https://github.com/pharo-project/pharo/pull/15225 # RPackage - Rpackage: Extension methods should know their package #15212 https://github.com/pharo-project/pharo/pull/15212 - Introduce Behavior>>#definedSelectors and #definedMethods and add tests #15264 https://github.com/pharo-project/pharo/pull/15264 - Simplify ClassRemoved and MethodRemoved now that packages are preserved #15262 https://github.com/pharo-project/pharo/pull/15262 - Make class know its package tag instead of its category #15239 https://github.com/pharo-project/pharo/pull/15239 - Ring-Definitions: Remove concept of categories #15235 https://github.com/pharo-project/pharo/pull/15235 - Deprecate Class>>#category: and ShiftClassBuilder>>#category(:) #15237 https://github.com/pharo-project/pharo/pull/15237 - Packages: Do some cleanings #15219 https://github.com/pharo-project/pharo/pull/15219 # Fluid Class Definitions - Fix more deprecated messages in traits #15275 https://github.com/pharo-project/pharo/pull/15275 # Compiler - Opal: Add concept of prior methods #15291 https://github.com/pharo-project/pharo/pull/15291 - Fix trait method recompilation and clean #15263 https://github.com/pharo-project/pharo/pull/15263 - Opal: Make plugins no-op to not crash during the loading #15289 https://github.com/pharo-project/pharo/pull/15289 # Undeclared - Introduce a specific collection for undeclareds #15271 https://github.com/pharo-project/pharo/pull/15271 - Ensure no association is added to the system dictionary #15241 https://github.com/pharo-project/pharo/pull/15241 # Tests - Clean traits tests #15243 https://github.com/pharo-project/pharo/pull/15243 - Release test: Add test that all packages are defined in a baseline #15273 https://github.com/pharo-project/pharo/pull/15273 - updates the #manuallyResolvedDependencies #15286 https://github.com/pharo-project/pharo/pull/15286 - Remove some classes generated by tests #15272 https://github.com/pharo-project/pharo/pull/15272 - Fix one more test cleanup #15248 https://github.com/pharo-project/pharo/pull/15248 - Improve cleanups of tests #15244 https://github.com/pharo-project/pharo/pull/15244 - Improve tests logs #15242 https://github.com/pharo-project/pharo/pull/15242 - Improve logs while running tests #15217 https://github.com/pharo-project/pharo/pull/15217 # Cleanup - Ring: Simplify Ring-Definitions/Ring-Definitions-Containers #15230 https://github.com/pharo-project/pharo/pull/15230 - add trace to #iconNamed:ifNone: to log all old names #15200 https://github.com/pharo-project/pharo/pull/15200 - Fix some icon references #15276 https://github.com/pharo-project/pharo/pull/15276 - Remove unused Calypso-SystemPlugins-Monticello-Queries #15270 https://github.com/pharo-project/pharo/pull/15270 - Merge Metacello-PharoExtensions into Metacello-Base #15215 https://github.com/pharo-project/pharo/pull/15215 - Calypso: Simplify ClyExtendingPackagesQuery adn clean package API #15221 https://github.com/pharo-project/pharo/pull/15221 # Trivial - Fix some deprecations #15238 https://github.com/pharo-project/pharo/pull/15238 - Cleanup: do not use notNil in Ring #15250 https://github.com/pharo-project/pharo/pull/15250 - Cleanup: usage of #notNil in Morphic #15260 https://github.com/pharo-project/pharo/pull/15260 - Cleanup: usage of #notNil in UnifiedFFI #15256 https://github.com/pharo-project/pharo/pull/15256 - Cleanup: Usage of notNil in Kernel #15269 https://github.com/pharo-project/pharo/pull/15269 - Cleanup: usage of #notNil in Athens, Graphics, Polymorph, FreeType, EmbeddedFreeType and FormCanvas #15267 https://github.com/pharo-project/pharo/pull/15267

Nov. 10, 2023

Re: [Pharo-vm] Stack overflow support?

by David Mason

Tail-call-elimination is certainly not a total solution to the stack overflow problem. However it would catch the `foo ^ self foo` example. And it would reduce he demand for stack. I like what Sven mentioned about a resumable exception. You could incrementally expand your stack in your application. As for debugging, it's not clear how much of a problem the lack of a return context actually is. We won't know until TCE is implemented and we get some experience with it. For the record Zag Smalltalk is doing TCE. ../Dave On Thu, 9 Nov 2023 at 09:18, Guille Polito <guillermo.polito(a)inria.fr> wrote: > > > El 9 nov. 2023, a las 13:35, David Mason <dmason(a)torontomu.ca> escribiÃ³: > > Tail call elimination would reduce stack usage significantly. See > Tail Call Elimination in OpenSmalltalk, Ralston & Mason, IWST2019 > > This really should be part of any stack solution. Particularly people > coming from a functional language background are likely to use it. Also > double dispatch would benefit significantly. > > > Yes, I agree that tail-call elimination will largely improve performance. > But it will also harm debugging unless there is a way to reconstruct the > stack when there is a problem, which I suspect could be very expensive. > > Anyways, this solves a performance problem, not the recursion problem that > motivates my mail, right? > > > ../Dave > > On Thu, 9 Nov 2023 at 04:38, Guillermo Polito <guillermopolito(a)gmail.com> > wrote: > >> Hi all, >> >> We started (with many interruptions over the last months) working a bit >> with Stephane on understanding what is the (positive and negative) impact >> of stack-overflow support in Pharo. >> The key idea is that if a process consumes too much stack (potentially >> because of an infinite recursion) then the process should stop with an >> exception. >> >> ## Why we want better stack consumption control >> >> This idea comes up to solve issues that are pretty common and hit >> especially newbies. >> For example, imagine you accidentally write an accessor such as >> >> ``` >> A >> foo >> ^ self foo >> ``` >> >> Students do this all the time, and Iâve also seen it in experienced >> people who go too fast :). >> More importantly, such recursions could happen also with not-so-obvious >> indirect recursions (a sends b, b sends c, c sends a), and these could hit >> anybody. >> >> This is aggravated because the current execution model allows us to have >> infinite stacks âmeaning: limited by available memory only. >> This is indeed a nice feature for many use cases but it has its own >> drawbacks when one of these kind of recursions are hit: >> - code just loops forever taking space in the stack >> - when there is no more stack space, context objects are created and >> moved to the heap >> - but those contexts are strongly held, so they are never GCed and take >> up extra space >> - even worse! they are there adding more work to the GC every time and >> making the GC run more often looking for space that is not there >> >> ## Why Ctrl-dot does not always work >> >> Of course, super users know there is this âCtrl dotâ hidden feature that >> should help you recover from this. >> First, let's take out of the equation that this is only known by super >> users. >> Now, in this situation, when Ctrl-dot is hit it will trigger a handler >> that suspends the problematic process and opens a debugger on it. >> But it could happen that, >> - the stack is so big that the debugger is very sluggish (best-case >> scenario) >> - the VM is just flooded doing GCs so maybe the Ctrl dot event does not >> even arrive at Pharo or the trigger >> - if the recursion is hit when printing an object (which is more common >> than you could imagine), opening the debugger could trigger a new recursion >> and never give back the control to the user >> >> ## What are we working on >> >> The main idea here is: Can we have a simple and efficient way to prevent >> such kinds of situations? >> >> After many discussions around detecting recursion, we kinda arrived at >> the simple solution of just detecting a stack overflow. >> The solution is easy to understand (because itâs like other languages >> work) and easy to implement because there is already support for that. >> But this leaves open two questions: >> - what happens when people want to use the âinfinite stackâ feature? >> - when should a process stack overflow? What is a sensitive default >> value? >> >> Our draft implementation here >> https://github.com/pharo-project/pharo-vm/pull/710 does the following to >> cope with this: >> - we can now parametrize the size of the stack (of each stack page to be >> more accurate) when the VM starts up >> - the stack overflow check can be disabled per process >> >> We also are running experiments to see what could be a sensitive stack >> size for our normal usages. Here, for example, we ran almost all test cases >> in Pharo separately (one suite per line below), and we observed how many >> tests broke (x-axis) with different stack sizes (y-axis). >> Here we see that most test suites require at least 20-24k to run >> properly, some go up to 36k of stack before converging (i.e., the number of >> broken tests does not change). >> >> <ImagenPegada-10.tiff> >> >> Youâll notice in the graph that There are some scenarios that break all >> the time. This is because exception handling itself is recursive and may >> produce more stack overflows depending on the size of the stack between the >> exception and the exception handler. >> So some more work is still required, mostly changing Pharo libraries to >> properly support this. For example: >> - should tests run in a fresh process with a fresh stack? >> - should the exception mechanism use less recursion? >> - resumable exceptions add stack pressure because they do not âunstackâ >> until the exception is finally handled, meaning that the stack used by >> exception handling just adds up to the stack of the original code, can we >> do better here? >> >> Probably there are more interesting questions here, thatâs the âwhy" >> behind this email. >> Iâm interested in opinions and scenarios you may come up with that should >> be taken into account. >> >> Cheers, >> Guille >> > _______________________________________________ > Pharo-vm mailing list -- pharo-vm(a)lists.pharo.org > To unsubscribe send an email to pharo-vm-leave(a)lists.pharo.org > > >

Nov. 9, 2023

Re: [Pharo-vm] Stack overflow support?

by Guille Polito

> El 9 nov. 2023, a las 13:35, David Mason <dmason(a)torontomu.ca> escribiÃ³: > > Tail call elimination would reduce stack usage significantly. See > Tail Call Elimination in OpenSmalltalk, Ralston & Mason, IWST2019 > > This really should be part of any stack solution. Particularly people coming from a functional language background are likely to use it. Also double dispatch would benefit significantly. Yes, I agree that tail-call elimination will largely improve performance. But it will also harm debugging unless there is a way to reconstruct the stack when there is a problem, which I suspect could be very expensive. Anyways, this solves a performance problem, not the recursion problem that motivates my mail, right? > > ../Dave > > On Thu, 9 Nov 2023 at 04:38, Guillermo Polito <guillermopolito(a)gmail.com <mailto:guillermopolito@gmail.com>> wrote: > Hi all, > > We started (with many interruptions over the last months) working a bit with Stephane on understanding what is the (positive and negative) impact of stack-overflow support in Pharo. > The key idea is that if a process consumes too much stack (potentially because of an infinite recursion) then the process should stop with an exception. > > ## Why we want better stack consumption control > > This idea comes up to solve issues that are pretty common and hit especially newbies. > For example, imagine you accidentally write an accessor such as > > ``` > A >> foo > ^ self foo > ``` > > Students do this all the time, and Iâve also seen it in experienced people who go too fast :). > More importantly, such recursions could happen also with not-so-obvious indirect recursions (a sends b, b sends c, c sends a), and these could hit anybody. > > This is aggravated because the current execution model allows us to have infinite stacks âmeaning: limited by available memory only. > This is indeed a nice feature for many use cases but it has its own drawbacks when one of these kind of recursions are hit: > - code just loops forever taking space in the stack > - when there is no more stack space, context objects are created and moved to the heap > - but those contexts are strongly held, so they are never GCed and take up extra space > - even worse! they are there adding more work to the GC every time and making the GC run more often looking for space that is not there > > ## Why Ctrl-dot does not always work > > Of course, super users know there is this âCtrl dotâ hidden feature that should help you recover from this. > First, let's take out of the equation that this is only known by super users. > Now, in this situation, when Ctrl-dot is hit it will trigger a handler that suspends the problematic process and opens a debugger on it. > But it could happen that, > - the stack is so big that the debugger is very sluggish (best-case scenario) > - the VM is just flooded doing GCs so maybe the Ctrl dot event does not even arrive at Pharo or the trigger > - if the recursion is hit when printing an object (which is more common than you could imagine), opening the debugger could trigger a new recursion and never give back the control to the user > > ## What are we working on > > The main idea here is: Can we have a simple and efficient way to prevent such kinds of situations? > > After many discussions around detecting recursion, we kinda arrived at the simple solution of just detecting a stack overflow. > The solution is easy to understand (because itâs like other languages work) and easy to implement because there is already support for that. > But this leaves open two questions: > - what happens when people want to use the âinfinite stackâ feature? > - when should a process stack overflow? What is a sensitive default value? > > Our draft implementation here https://github.com/pharo-project/pharo-vm/pull/710 <https://github.com/pharo-project/pharo-vm/pull/710> does the following to cope with this: > - we can now parametrize the size of the stack (of each stack page to be more accurate) when the VM starts up > - the stack overflow check can be disabled per process > > We also are running experiments to see what could be a sensitive stack size for our normal usages. Here, for example, we ran almost all test cases in Pharo separately (one suite per line below), and we observed how many tests broke (x-axis) with different stack sizes (y-axis). > Here we see that most test suites require at least 20-24k to run properly, some go up to 36k of stack before converging (i.e., the number of broken tests does not change). > > <ImagenPegada-10.tiff> > > Youâll notice in the graph that There are some scenarios that break all the time. This is because exception handling itself is recursive and may produce more stack overflows depending on the size of the stack between the exception and the exception handler. > So some more work is still required, mostly changing Pharo libraries to properly support this. For example: > - should tests run in a fresh process with a fresh stack? > - should the exception mechanism use less recursion? > - resumable exceptions add stack pressure because they do not âunstackâ until the exception is finally handled, meaning that the stack used by exception handling just adds up to the stack of the original code, can we do better here? > > Probably there are more interesting questions here, thatâs the âwhy" behind this email. > Iâm interested in opinions and scenarios you may come up with that should be taken into account. > > Cheers, > Guille > _______________________________________________ > Pharo-vm mailing list -- pharo-vm(a)lists.pharo.org > To unsubscribe send an email to pharo-vm-leave(a)lists.pharo.org

Nov. 9, 2023